Horangi 4 - Ko-HalluLens (WikiQA): leaderboard
Ko-HalluLens WikiQA, Korean questions about Wikipedia facts from the HalluLens suite, as run by the Horangi Korean LLM Leaderboard 4 (W&B Korea). Correct-answer rate (%), higher is better.
Metric: Correct answer rate (%). Source: horangi.ai. Saturation forecast: Around August 2027. 105 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 64 | #22 (GPT-5.6 Sol) |
| 2 | Gemini 3.1 Pro (Preview) (High) | 64 | #64 (Gemini 3.1 Pro (Preview)) |
| 3 | GPT-5.5 (xHigh) | 63 | #33 (GPT-5.5) |
| 4 | Gemini 3 Pro (Preview) (High) | 62 | #71 (Gemini 3 Pro (Preview)) |
| 5 | Grok 4.5 | 60 | #54 |
| 6 | Gemini 3.5 Flash | 59 | #66 |
| 7 | Gemini 3 Flash (Preview) | 58 | #90 |
| 8 | Claude Fable 5 (High) | 55 | #20 (Claude Fable 5) |
| 9 | GPT-5.4 (2026-03-05) (xHigh) | 54 | #63 (GPT-5.4 (2026-03-05)) |
| 10 | GPT-5.6 Luna (Max) | 52 | #79 (GPT-5.6 Luna) |
| 11 | GPT-5 | 51 | #105 |
| 12 | Gemini 3.1 Flash Lite (Preview) | 50 | #201 |
| 13 | GPT-5.1 (2025-11-13) | 48 | #107 |
| 14 | GPT-5.6 Terra (Max) | 48 | #40 (GPT-5.6 Terra) |
| 15 | GPT-4.1 | 45 | #258 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=horangi-4-ko-hallulens-wikiqa · How It Works · Data refreshed daily, snapshot 2026-11-05.